Search

The traditional matrix decomposition model cannot fully explored the intrinsic relationship between the user and the object in the mean, bias and characteristics. This paper proposed a fitting matrix model to improve the prediction performance by constructing the user and the item matrix to represent the characteristics of the user and the item respectively. The matrix decomposition model had the advantage of accuracy in the field of recommender system, but the gradient descent method, which was the most popular method to train parameters of model, had a slow convergence speed. To resolve the above defects, this paper considered to accelerate the convergence speed using the convergence of quasi Newton method, and named the proposed algorithm as fitting matrix and two orders fusion iterative (FAST) algorithm. The experimental results show that the FAST algorithm is better than the traditional non negative matrix decomposition (NMF), singular value matrix decomposition (SVD), and the regularized singular value matrix decomposition (RSVD). FAST algorithm has a decrease with regard to the mean absolute error (MAE) and the root mean square error (RMSE), and has a significant improvement in the iterative efficiency, which alleviates the problem that the accuracy is difficult to balance with the efficiency of the iteration. [ABSTRACT FROM AUTHOR]

Traditional network optimization problems are always solved by the dual gradient descent algorithm, which although can be implemented in a distributed manner, has a slow convergence rate. The accelerated dual descent (ADD) algorithms improve the convergence rate of dual gradient descent algorithm through distributed computation of approximated Newton steps. But with the uncertainty of communication networks, the convergence of the algorithm cannot be guaranteed under uncertain constraints. Based on this, this paper proposed a stochastic version of ADD algorithm to solve the network optimization problems under uncertainty. It proved theoretically that the stochastic ADD algorithms could almost surely converge to an error neighborhood of the optimal when the mean square error of the uncertainty was bounded, and gave a more strict constraint of uncertainty, can exactly almost surely converge to the optimal point. Numerical results show that the stochastic ADD algorithms converge in two orders of magnitude less iteration than the stochastic gradient descent algorithms. [ABSTRACT FROM AUTHOR]

The road is unstable under emergency conditions (like earthquakes, landslides, floods, etc ), accident waiting to happen. To solve problem of logistics distribution under emergency conditions, based on the study of traditional ant colony algorithm in solving VRP (vehicle routing problem), this paper tried to improve VRP mathematical model by adding a dynamic road factor. It gave a mathematical mode of ant colony algorithm to solve improved vehicle routing problem, and used the actual latitude and longitude of 21 cities in Shaanxi Province as an emergency analog cargo demanding places and warehouse locations to make a system simulation. It analyzed the parameters which had impact on improved algorithm convergence. Theoretical studies and MATLAB simulation proves that the improved algorithm can effectively solve the actual situation of emergency logistics distribution. [ABSTRACT FROM AUTHOR]

Because the calculation of searching spectral peak process is heavy, MUSIC algorithms can't meet the requirement of high real-time. To solve this problem, this paper proposed an improved population genetic algorithms (IPGA). In the algorithms, it crossovered the best chromosome called queen among the current population with drones selected according to a certain crossover probability, which enhanced the exploitation of searching global optimum. Te algorithm used the strategy by stages to adjust the dynamic population scale, at the same time, it adopted a local search to randomly produce more parents around the secondary best chromosome, so as to escape the premature and increase the exploration ability. The algorithm used the strategy by stages to adjust the dynamic random population seal. The random population size changing gradually, guarantees not only the diversity of population, but also improve the ability to overcome the problem of premature convergence effectively. Compared genetic algorithms and the traversal algorithms, genetic algorithms can largely reduce the computation of the MUSIC spectrum peak search. [ABSTRACT FROM AUTHOR]

This paper proposed a Web log mining technology orienting the demand of mining for TB-level log files. It designed and implemented a genus of GA which deployed on the Yarn using the MapReduce programing model to combine the GA with the Yarn effectively. In addition, took the Round-Robin strategy instead of the hash method during the Partition stage of MapReduce in order to distribute the gene among the sub populations evenly, increased the populations' rate of convergence and validity of result, and balanced the workload amid all of the nodes at the same time. Finally, the test shows that the rate of the validate result using the technology, which improves the efficiency by 33%, the average accuracy is more than 93%. [ABSTRACT FROM AUTHOR]

This paper put forward a new hybrid algorithm taking advantage of bacteria foraging optimization algorithm and the idea of catfish effect to transform its individual evolutionary strategies. It used the hybrid algorithm to test four classical functions and modelled Wiener simulation. The results show its superiority in the overall searching ability, convergence rate and solution precision than other improved PSO. [ABSTRACT FROM AUTHOR]

Concerning the problem that the independent adjusting strategy of inertia weight and learning factor reduces evolution uniformity of particle swarm optimization (PSO) algorithm, and cannot adapt to the complex nonlinear optimization problems, this paper proposed a new PSO algorithm with learning factor controlled by inertia weight function. This strategy could effectively enhance the interaction of inertia weight and learning factor so as to balance the global exploration and local exploitation and preferably lead particles to search globally optimal solution. Based on that, it introduced the time factor, which treated as a linear function of inertia weight, in order to further improve the local development ability and convergence speed of iteration in the late. Aiming at the conflict between the convergence and the diversity of PSO, it proposed an strategy of boundary restrictions and speed rebound, which could avoid particles flying off resulting in a decrease of species diversity and made particles converge to global optimum quickly. The optimization analysis on benchmark test functions and the comparison with other PSO algorithm indicates that the algorithm balances individual and colonial learning ability of particles and improves optimization ability and the accuracy of convergence. [ABSTRACT FROM AUTHOR]

Focusing on the problem of high efficiency resource scheduling in cloud computing environment, since current re source scheduling algorithm had been less given consideration to the shortest completion time and the least cost of the services, this paper designed the fitness function which could comprehensive response the time and cost, and proposed an estimation of distribution-shuffled frog leaping algorithm (EDSFLA). This algorithm redefined the evolutional operators of shuffled frog lea ping algorithm (SFLA), utilizing the cross operators of genetic algorithm, targeted in applying to scheduling problem with integer-coded. EDSFLA introduced evolutionary strategy of estimation of distribution to break the confine of search pattern in the standard SFLA, ad made the learning ability of this algorithm more comprehensive. Experimental results show that this algorithm has more capability in convergence and searching ability compared with the standard SFLA and EDA in solving cloud resource scheduling problems. [ABSTRACT FROM AUTHOR]

After analyzing the disadvantages of slow convergence in the late process and local extreme of particle swarm optimization, this paper proposed a simplified PSO clustering algorithm based on granules and maximum distances product method. By adopting tactics of linearly decreasing weight and random distribution, adding the extremum disturbance operator, optimizing the individual extremum of particles, the PSO algorithm could quickly converge to global optimal. Ten the improved PSO algorithm combined with K-means algorithm, to solve poor clustering effect and instability of K-means algorithm caused by random initial clustering center. Experiment results show that this algorithm has higher accuracy, faster convergence rate and stronger stability. [ABSTRACT FROM AUTHOR]